Pith. sign in

REVIEW 3 major objections 5 minor 57 references

ASANet: Asymmetric Semantic Aligning Network for RGB and SAR image land cover classification

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read ASANet treats RGB and SAR as asymmetric partners and reports the best land cover classification on three multimodal datasets.

desk verdict A cleanly-written RGB-SAR fusion paper with a useful new cloudy benchmark, but the SOTA claim is not statistically supported by single-run margins that fall inside run-to-run noise. read the letter →

arxiv 2412.02044 v1 pith:DZLXMRDN submitted 2024-12-03 eess.IV cs.CV

Pan Zhang , Baochai Peng , Chaoran Lu , Quanjin Huang This is my paper · ORCID
classification eess.IVcs.CV
keywords landcoverclassificationmultimodalfusionRGB-SARsemanticsegmentationasymmetricattentiondifferentialfeaturesremotesensing
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that RGB and SAR imagery should not be treated as symmetric partners in a fusion network. It proposes ASANet, whose Semantic Focusing Module derives per-modality channel weights from the pixel-wise difference between the two modalities, and whose Cascade Fusion Module calibrates channel and spatial features before fusing them. On three RGB-SAR datasets, including a newly collected Pearl River $\Delta$ dataset with real cloud cover, ASANet reports higher mean Intersection over Union than six existing fusion networks, with gains of 1.21% to 17.69% on the new dataset. The claim, read sympathetically, is that letting each modality keep its own focus before aligning the two extracts complementary land-cover cues that symmetric attention fusion misses.

What carries the argument

The load-bearing object is the pair of modules designed around asymmetry. The Semantic Focusing Module (SFM) computes signed difference maps $F^r_\Delta = F_{RGB} - F_{SAR}$ and $F^s_\Delta = F_{SAR} - F_{RGB}$, pools them with global max-pooling, refines them with convolutional channel-wise perceptrons, and gates each original branch with its own sigmoid weight. The Cascade Fusion Module (CFM) concatenates the gated branches, extracts channel attention with two non-shared MLPs after global average pooling, then extracts spatial attention with two softmax-weighted convolutions, and sums the two weighted feature maps. SFM is what makes the fusion asymmetric; CFM is what selects and aligns the recalibrated features.

What would settle it

Run ASANet on PIE-RGB-SAR with the SAR branch deliberately shifted by one to three pixels relative to the RGB branch and compare mIoU with the aligned model: if the signed differences encode true complementary semantics, this small misregistration should produce a sharp, systematic drop, whereas if the gains come from the surrounding attention machinery, the drop should be small and unsystematic. Replacing the subtraction in Eq. (1) with absolute difference or concatenation while keeping the rest of the network fixed would also directly test whether the signed difference is the load-bearing operation.

Watch

Extended reading notes

Core claim

The paper claims that a network built on asymmetric feature interaction outperforms symmetric multimodal fusion for RGB-SAR land cover classification. The Semantic Focusing Module computes differential feature maps by subtracting the two modality feature maps in opposite directions, converts those differences into channel weights through global max-pooling and convolutions, and applies the weights to the original branches so that each modality keeps distinctive information. The Cascade Fusion Module then concatenates the recalibrated branches, applies separate channel and spatial attention, and adds the weighted features. Trained with two ConvNeXtV2-tiny backbones and an UPerNet decoder, ASANet reaches 78.31% mIoU on PIE-RGB-SAR, 94.48% on DDHR-SK, and 56.11% on WHU-OPT-SAR, the best among all compared methods.

Load-bearing premise

The entire asymmetry mechanism rests on the pixel-wise subtraction between RGB and SAR feature maps being a meaningful semantic comparison, which requires accurate co-registration and matching scale between the two images; because the acquisitions are not simultaneous, any misalignment or land-cover change would turn the difference features into alignment artifacts.

Editorial extensions

If this is right

  • On the new PIE-RGB-SAR dataset, ASANet improves mIoU by 1.21% to 17.69% over six existing multimodal segmentation networks.
  • On DDHR-SK and WHU-OPT-SAR, ASANet reaches 94.48% and 56.11% mIoU, the highest scores among the compared methods.
  • The SFM keeps the two branches visually distinct after feature interaction, whereas a symmetric interaction like CMX makes them look similar, and ASANet's fused features show fewer fragmented false-detection regions.
  • Replacing pixel-wise addition with the Cascade Fusion Module contributes more to accuracy than simply adding SFM to the baseline, especially for road, water, and farmland categories.
  • ASANet runs at 48.7 FPS for 256x256 inputs on a Quadro RTX 8000, 15.2 FPS faster than the next-best model CMX.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The signed-difference mechanism in SFM may transfer to other modality pairs with different imaging physics, such as optical and LiDAR or multi-temporal SAR, where one modality is noisy or sparse; that transfer is an extension, not tested in this paper.
  • The largest gains on road and water suggest SFM acts partly as a change detector between the two acquisitions; an ablation that replaces signed subtraction with absolute difference or concatenation would show whether the sign of the difference carries the signal.
  • The stage ablation indicates most of the benefit comes from stage-2 and stage-3 fusion, so a lightweight variant using SFM and CFM only at those stages could retain much of the gain at higher speed; the paper reports all-stage fusion as best, but this narrower configuration is a natural test.
  • Because PIE-RGB-SAR contains real cloud cover and non-simultaneous RGB-SAR acquisitions, it may be a more demanding benchmark for fusion robustness than datasets with simulated clouds; evaluating other fusion methods on it beyond the six tested would clarify how general the gap is.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This paper presents ASANet, a dual-branch network for land cover classification from RGB and SAR images. It introduces a Semantic Focusing Module (SFM) that computes per-branch channel attention from pixel-wise RGB-SAR difference features, and a Cascade Fusion Module (CFM) that applies channel- and spatial-wise attention to fuse the two modalities. The authors also collect and annotate a new dataset, PIE-RGB-SAR, and report experiments on this dataset plus DDHR-SK and WHU-OPT-SAR, claiming state-of-the-art mIoU performance and a favorable FPS/complexity trade-off.

Significance. If the reported results hold up, the paper makes a useful contribution by showing that asymmetric, modality-specific attention can outperform symmetric fusion in RGB-SAR classification, and the new PIE-RGB-SAR dataset could be a valuable benchmark. The public release of source code is a strength. However, the currently reported evidence does not establish the SOTA claim with confidence, and the architectural description contains a dimension inconsistency that must be resolved.

major comments (3)
  1. [§4.2, Tables 2–5] All accuracy results are from single training runs, and the margins over the strongest baseline (CMX) are 1.21, 0.16, and 0.43 mIoU on the three datasets. With batch size 4, data augmentation, and random initialization, run-to-run variation on such datasets is typically on the order of 0.5 mIoU points or more, so the two smaller margins are within noise. Please run each method with at least 3–5 seeds and report mean and standard deviation, and state whether the improvements are consistent across seeds. Also, since the paper reports the checkpoint with the best validation mIoU, please discuss the potential selection bias from this protocol.
  2. [§3.3, Eq. (5)] The notation is internally inconsistent: the text states that z ∈ R^{1×H×W}, that F_c^2 has an output dimension of c, and that F3 and F4 have input/output dimensions C. If c ≠ 1 and c ≠ C, these layers cannot be applied to z as described. Please clarify the actual channel dimensions, the role of the compression factor c, and correct the equations accordingly.
  3. [§4.1.1 and Eq. (1)] The SFM computes pixel-wise differences between RGB and SAR feature maps. The dataset description states that the RGB (≈0.5 m) and SAR (3 m) images were resampled and aligned with PIE-Basic and were not acquired simultaneously. Residual misregistration or land-cover change between the two acquisitions could make these difference features encode alignment artifacts. Please provide a quantitative alignment error estimate or a sensitivity experiment (e.g., with artificial shifts) to show that the SFM is robust to realistic misregistration, or discuss why the global-max-pooling operator makes this concern moot.
minor comments (5)
  1. [§4.1.1] The PIE-RGB-SAR dataset is advertised as a contribution, but no public download link or data availability statement is given; please add one or state the terms under which the dataset can be accessed.
  2. [Table 6] The checkmark columns are not clearly labeled; the text refers to rows 3, 5, and 6 as PWA, SFM+PWA, and SFM+CFM, but as printed, it is hard to decode which module is active in each row. Please reformat the table with explicit column headers and marks.
  3. [Abstract and §1] The phrase 'improvement in mIoU ranging from 1.21% to 17.69%' could be read as the range over all compared methods; please clarify that 1.21% is the gain over CMX and 17.69% is the gain over FuseNet.
  4. [§2.2] The citation '(Gao et al., 2023)' is given without a method name; please provide a brief description or name.
  5. [General] There are minor grammatical errors, e.g., 'The source code are available', and some awkward sentences in Section 5; a careful language edit is recommended.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: ASANet is trained end-to-end on held-out validation partitions and the reported mIoU gains are empirical outcomes, not quantities derived from the model equations.

full rationale

The paper's derivation chain is an architectural proposal followed by empirical evaluation. The SFM (Eqs. 1-3) and CFM (Eqs. 4-6) define differentiable operations whose parameters are learned by minimizing a segmentation loss; none of these equations contains the reported mIoU values or is fitted to them. The claim that ASANet 'outperforms other methods' is supported by held-out validation metrics in Tables 2-5, and the ablations in Tables 6-7 compare architectural variants on the same evaluation protocol. No fitted parameter is renamed as a prediction, and no result is justified solely by a self-citation: the cited backbones, decoders, attention modules, and datasets are external prior work, while the newly introduced PIE-RGB-SAR dataset is a benchmark contribution whose labels are independent of the model. The absence of repeated runs and error bars around margins such as +0.16 and +0.43 mIoU is a legitimate statistical-robustness concern, but it concerns evidence strength rather than circularity.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

ASANet is an empirical network, not a derivation. Its central claim rests on dataset alignment and annotation quality, transfer of pretrained weights, and the stability of single-run validation metrics. SFM and CFM are network modules, not invented entities, and the only hand-chosen quantities are architecture and training hyperparameters, listed above.

free parameters (4)
  • ImageNet22k pretrained weights = ConvNeXtV2-tiny pretrained on ImageNet22k
    Used to initialize both branches; the transfer from natural images to SAR is assumed and not measured, and branch symmetry is broken only after pretraining.
  • SFM channel reduction dimension = max(32, C/16)
    Chosen by hand in Eq. 2; controls the capacity of the differential channel reweighting and is not derived from data.
  • CFM compressed channel dimension = C after convolutions on concatenated 2C input (text is ambiguous)
    The text does not clearly specify the intermediate channel count; the architecture as printed in Eq. 5 cannot be implemented exactly.
  • Training schedule = 80k iterations, batch size 8, AdamW lr 1e-4, best validation checkpoint
    Hyperparameters selected by hand; the reported accuracies correspond to the best validation checkpoint, which can inflate single-run performance.
assumptions (4)
  • domain assumption The two modality images are accurately co-registered and resampled so pixel-wise subtraction in Eq. 1 is semantically meaningful.
    Section 4.1.1 describes resampling with PIE-Basic, but the RGB (Google, about 0.5 m) and SAR (GF3, 3 m) images were not acquired simultaneously and misregistration is not quantified.
  • domain assumption PIE-RGB-SAR annotations are correct and consistent despite category imbalance and non-simultaneous acquisition.
    Section 4.1.1 states that RGB is the primary source and SAR is complementary; label quality is asserted but no annotation protocol, inter-annotator agreement, or quality checks are reported.
  • domain assumption ImageNet22k pretrained ConvNeXtV2 weights transfer to both optical and SAR remote sensing.
    Section 4.2 adopts ImageNet22k pretraining for both backbones; SAR's different image statistics make this nontrivial and it is not validated independently.
  • domain assumption Validation mIoU from a single train/validation split is a stable estimator of model ranking.
    Each model is evaluated once in Tables 2 to 5, with no repeated seeds or error bars, yet the SOTA claim depends on small differences in mIoU.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ASANet: Asymmetric Semantic Aligning Network for RGB and SAR image land cover classification." pith.science (2026). https://pith.science/paper/DZLXMRDN

@misc{pith2026241202044,
  author       = {Pith},
  title        = {Pith review of: ASANet: Asymmetric Semantic Aligning Network for RGB and SAR image land cover classification},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DZLXMRDN}},
  note         = {Machine review of arXiv:2412.02044}
}
read the original abstract

Synthetic Aperture Radar (SAR) images have proven to be a valuable cue for multimodal Land Cover Classification (LCC) when combined with RGB images. Most existing studies on cross-modal fusion assume that consistent feature information is necessary between the two modalities, and as a result, they construct networks without adequately addressing the unique characteristics of each modality. In this paper, we propose a novel architecture, named the Asymmetric Semantic Aligning Network (ASANet), which introduces asymmetry at the feature level to address the issue that multi-modal architectures frequently fail to fully utilize complementary features. The core of this network is the Semantic Focusing Module (SFM), which explicitly calculates differential weights for each modality to account for the modality-specific features. Furthermore, ASANet incorporates a Cascade Fusion Module (CFM), which delves deeper into channel and spatial representations to efficiently select features from the two modalities for fusion. Through the collaborative effort of these two modules, the proposed ASANet effectively learns feature correlations between the two modalities and eliminates noise caused by feature differences. Comprehensive experiments demonstrate that ASANet achieves excellent performance on three multimodal datasets. Additionally, we have established a new RGB-SAR multimodal dataset, on which our ASANet outperforms other mainstream methods with improvements ranging from 1.21% to 17.69%. The ASANet runs at 48.7 frames per second (FPS) when the input image is 256x256 pixels. The source code are available at https://github.com/whu-pzhang/ASANet

Figures

Figures reproduced from arXiv: 2412.02044 by the authors.

Figure 1
Figure 1. Comparison of different fusion frameworks: Features are categorized based on whether they engage in direct interaction within the network, being classified into two types: a) Indirect Interaction Segmentation Framework and b) Direct Interaction Segmentation Framework. a) Feature Indirect Interaction Framework: This framework refers to a segmentation network that lacks direct feature interaction between two modalitie… view at source ↗
Figure 2
Figure 2. Overall Network Structure of ASANet: The SFM concentrates on independently complementary features of [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Diagram of the structure of the SFM. Specifically, the process comprises three main steps. We denote the feature maps of the two modalities as FRGB ∈ R C×H×W and FSAR ∈ R C×H×W, respectively, where C, H, and W correspond to the number of channels, height, and width of the feature maps. Superscripts r (RGB) and s (SAR) are utilized to distinguish between the feature maps of the two modalities. Obtain differential fea… view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Diagram of the structure of the CFM. Suppose the input feature maps to the CFM are FRGB ∈ R C×H×W and FSAR ∈ R C×H×W, where C, H, and W represent the number of channels, height, and width of the feature maps, respectively. The superscripts r (RGB) and s (SAR) are used …
Figure 5
Figure 5. Figure 5: PIE-RGB-SAR dataset: RGB image on the left, SAR image in the middle, and ground truth image on the right. [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: DDHR-SK dataset: Original RGB image on the left, RGB image with simulated cloudy in the center, and SAR [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Part of the WHU-OPT-SAR dataset: RGB image on the left, SAR image in the center, and ground truth image [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: Visualization results of different methods on PIE-RGB-SAR dataset: four representative multimodal models. The visualization results of each model on the PIE-RGB-SAR dataset are shown in [PITH_FULL_IMAGE:figures/full_fig_p014_8.png]
Figure 9
Figure 9. Figure 9: Feature map visualization: (a) and (f) show the superposition ground truth of RGB and SAR images, re [PITH_FULL_IMAGE:figures/full_fig_p015_9.png]
Figure 10
Figure 10. Figure 10: Visualization results of different methods on DDHR-SK dataset: four representative multimodal models [PITH_FULL_IMAGE:figures/full_fig_p016_10.png]
Figure 11
Figure 11. Figure 11: Visualization results of different methods on WHU-OPT-SAR dataset: four representative multimodal models. The results of different models on the WHU-OPT-SAR dataset are shown in [PITH_FULL_IMAGE:figures/full_fig_p017_11.png]
Figure 12
Figure 12. Figure 12: Visualization results of ablation experiments. [PITH_FULL_IMAGE:figures/full_fig_p019_12.png]
Figure 13
Figure 13. Figure 13: Overall image of the simulated cloudy scene: RGB on the left, Simulated RGB in the middle, and SAR on [PITH_FULL_IMAGE:figures/full_fig_p021_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

57 extracted references · 53 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in ":" * " " * FUNCTION f...

  2. [2]

    , author Lopatin, J

    author Araya-L \'o pez, R.A. , author Lopatin, J. , author Fassnacht, F.E. , author Hern \'a ndez, H.J. , year 2018 . title Monitoring andean high altitude wetlands in central chile with seasonal optical data: A comparison between worldview-2 and sentinel-2 imagery . journal ISPRS journal of photogrammetry and remote sensing volume 145 , pages 213--224

    work page 2018

  3. [3]

    , author Happ, P.N

    author Bermudez, J.D. , author Happ, P.N. , author Feitosa, R.Q. , author Oliveira, D.A. , year 2019 . title Synthesis of multispectral optical images from sar/optical multitemporal data using conditional generative adversarial networks . journal IEEE Geoscience and Remote Sensing Letters volume 16 , pages 1220--1224

    work page 2019

  4. [4]

    , author Jiang, L

    author Cao, Z. , author Jiang, L. , author Yue, P. , author Gong, J. , author Hu, X. , author Liu, S. , author Tan, H. , author Liu, C. , author Shangguan, B. , author Yu, D. , year 2023 . title A large scale training sample database system for intelligent interpretation of remote sensing imagery . journal Geo-Spatial Information Science , pages 1--20

    work page 2023

  5. [5]

    , author Lin, K.Y

    author Chen, X. , author Lin, K.Y. , author Wang, J. , author Wu, W. , author Qian, C. , author Li, H. , author Zeng, G. , year 2020 . title Bi-directional cross-modality feature propagation with separation-and-aggregation gate for rgb-d semantic segmentation , in: booktitle European Conference on Computer Vision , organization Springer . pp. pages 561--577

    work page 2020

  6. [6]

    , author Foody, G.M

    author Costa, H. , author Foody, G.M. , author Boyd, D.S. , year 2018 . title Supervised methods of image segmentation accuracy assessment in land cover mapping . journal Remote Sensing of Environment volume 205 , pages 338--351

    work page 2018

  7. [7]

    , author Xu, J

    author Gao, M. , author Xu, J. , author Yu, J. , author Dong, Q. , year 2023 . title Distilled heterogeneous feature alignment network for sar image semantic segmentation . journal IEEE Geoscience and Remote Sensing Letters

    work page 2023

  8. [8]

    , author Shengoku, H

    author Guo, Z. , author Shengoku, H. , author Wu, G. , author Chen, Q. , author Yuan, W. , author Shi, X. , author Shao, X. , author Xu, Y. , author Shibasaki, R. , year 2018 . title Semantic segmentation for urban planning maps based on u-net , in: booktitle IGARSS 2018-2018 IEEE International Geoscience and Remote Sensing Symposium , organization IEEE ....

    work page 2018

Show all 57 references
  1. [9]

    , author Wu, L

    author Guo, Z. , author Wu, L. , author Huang, Y. , author Guo, Z. , author Zhao, J. , author Li, N. , year 2022 . title Water-body segmentation for sar images: past, current, and future . journal Remote Sensing volume 14 , pages 1752

    2022

  2. [10]

    , author Li, Z

    author Hang, R. , author Li, Z. , author Ghamisi, P. , author Hong, D. , author Xia, G. , author Liu, Q. , year 2020 . title Classification of hyperspectral and lidar data using coupled cnns . journal IEEE Transactions on Geoscience and Remote Sensing volume 58 , pages 4939--4950

    2020

  3. [11]

    , author Ma, L

    author Hazirbas, C. , author Ma, L. , author Domokos, C. , author Cremers, D. , year 2017 . title Fusenet: Incorporating depth into semantic segmentation via fusion-based cnn architecture , in: booktitle Computer Vision--ACCV 2016: 13th Asian Conference on Computer Vision, Tai...

    2017

  4. [12]

    , author Sun, Y.J

    author He, T. , author Sun, Y.J. , author Xu, J.D. , author Wang, X.J. , author Hu, C.R. , year 2014 . title Enhanced land use/cover classification using support vector machines and fuzzy k-means clustering algorithms . journal Journal of Applied Remote Sensing volume 8 , page...

    2014

  5. [13]

    , author Yokoya, N

    author He, W. , author Yokoya, N. , year 2018 . title Multi-temporal sentinel-1 and-2 data fusion for optical image simulation . journal ISPRS International Journal of Geo-Information volume 7 , pages 389

    2018

  6. [14]

    , author Mourali, O

    author Hermessi, H. , author Mourali, O. , author Zagrouba, E. , year 2021 . title Multimodal medical image fusion review: Theoretical background and recent advances . journal Signal Processing volume 183 , pages 108036

    2021

  7. [15]

    , author Shen, L

    author Hu, J. , author Shen, L. , author Sun, G. , year 2018 . title Squeeze-and-excitation networks , in: booktitle Proceedings of the IEEE conference on computer vision and pattern recognition , pp. pages 7132--7141

    2018

  8. [16]

    , author Yang, K

    author Hu, X. , author Yang, K. , author Fei, L. , author Wang, K. , year 2019 . title Acnet: Attention based network to exploit complementary features for rgbd semantic segmentation , in: booktitle 2019 IEEE International Conference on Image Processing (ICIP) , organization I...

    2019

  9. [17]

    , author Xiang, Y

    author Kang, W. , author Xiang, Y. , author Wang, F. , author You, H. , year 2022 . title Cfnet: A cross fusion network for joint land cover classification using optical and sar images . journal IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing v...

    2022

  10. [18]

    , author Janusonis, E

    author Kazakeviciute-Januskeviciene, G. , author Janusonis, E. , author Bausys, R. , author Limba, T. , author Kiskis, M. , year 2020 . title Assessment of the segmentation of rgb remote sensing images: a subjective approach . journal Remote Sensing volume 12 , pages 4152

    2020

  11. [19]

    , author Hong, D

    author Li, J. , author Hong, D. , author Gao, L. , author Yao, J. , author Zheng, K. , author Zhang, B. , author Chanussot, J. , year 2022 a. title Deep learning in multimodal remote sensing data fusion: A comprehensive review . journal International Journal of Applied Earth O...

    2022

  12. [20]

    , author Zheng, S

    author Li, R. , author Zheng, S. , author Duan, C. , author Wang, L. , author Zhang, C. , year 2022 b. title Land cover classification from remote sensing images based on multi-scale fully convolutional network . journal Geo-spatial information science volume 25 , pages 278--294

    2022

  13. [21]

    , author Sun, K

    author Li, W. , author Sun, K. , author Li, W. , author Wei, J. , author Miao, S. , author Gao, S. , author Zhou, Q. , year 2023 a. title Aligning semantic distribution in fusing optical and sar images for land use classification . journal ISPRS Journal of Photogrammetry and R...

    2023

  14. [22]

    , author Zhang, G

    author Li, X. , author Zhang, G. , author Cui, H. , author Hou, S. , author Chen, Y. , author Li, Z. , author Li, H. , author Wang, H. , year 2023 b. title Progressive fusion learning: A multimodal joint segmentation framework for building extraction from optical and sar image...

    2023

  15. [23]

    , author Zhang, G

    author Li, X. , author Zhang, G. , author Cui, H. , author Hou, S. , author Wang, S. , author Li, X. , author Chen, Y. , author Li, Z. , author Zhang, L. , year 2022 c. title Mcanet: A joint semantic segmentation framework of optical and sar images for land use classification ...

    2022

  16. [24]

    , author Zhao, H

    author Li, X. , author Zhao, H. , author Wu, D. , author Liu, Q. , author Tang, R. , author Li, L. , author Xu, Z. , author Lyu, X. , year 2024 . title Slmfnet: Enhancing land cover classification of remote sensing images through selective attentions and multi-level feature fu...

    2024

  17. [25]

    , author Jiao, Q

    author Li, Z. , author Jiao, Q. , author Liu, L. , author Tang, H. , author Liu, T. , year 2014 . title Monitoring geologic hazards and vegetation recovery in the wenchuan earthquake region using aerial photography . journal ISPRS International Journal of Geo-Information volum...

    2014

  18. [26]

    , author He, C

    author Liu, X. , author He, C. , author Zhang, Q. , author Liao, M. , year 2019 . title Statistical convolutional neural network for land-cover classification from sar images . journal IEEE Geoscience and Remote Sensing Letters volume 17 , pages 1548--1552

    2019

  19. [27]

    , author Wu, Y

    author Liu, X. , author Wu, Y. , author Liang, W. , author Cao, Y. , author Li, M. , year 2022 . title High resolution sar image classification using global-local network structure based on vision transformer and cnn . journal IEEE Geoscience and Remote Sensing Letters volume ...

    2022

  20. [28]

    , year 2021

    author Lu, c. , year 2021 . title Geological Hazards Detection and Deformation Analysis in Mountainous Areas of Guizhou Based on Space-borne InSAR Technology . Master's thesis. Wuhan University

    2021

  21. [29]

    , author Zheng, C

    author Ma, A. , author Zheng, C. , author Wang, J. , author Zhong, Y. , year 2023 . title Domain adaptive land-cover classification via local consistency and global diversity . journal IEEE Transactions on Geoscience and Remote Sensing

    2023

  22. [30]

    , author Karaku s , O

    author Ma, W. , author Karaku s , O. , author Rosin, P.L. , year 2022 a. title Amm-fusenet: Attention-based multi-modal image fusion network for land cover mapping . journal Remote Sensing volume 14 , pages 4458

    2022

  23. [31]

    , author Wu, P

    author Ma, X. , author Wu, P. , author Wu, Y. , author Shen, H. , year 2017 . title A review on recent developments in fully polarimetric sar image despeckling . journal IEEE Journal of selected topics in applied earth observations and remote sensing volume 11 , pages 743--758

    2017

  24. [32]

    , author Zhang, X

    author Ma, X. , author Zhang, X. , author Pun, M.O. , year 2022 b. title A crossmodal multiscale fusion network for semantic segmentation of remote sensing data . journal IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing volume 15 , pages 3463--3474

    2022

  25. [33]

    , author Zhang, X

    author Ma, X. , author Zhang, X. , author Pun, M.O. , author Liu, M. , year 2024 . title A multilevel multimodal fusion transformer for remote sensing semantic segmentation . journal IEEE Transactions on Geoscience and Remote Sensing

    2024

  26. [34]

    , author Hor \'a k, J

    author Orl \' kov \'a , L. , author Hor \'a k, J. , year 2019 . title Land cover classification using sentinel-1 sar data , in: booktitle 2019 international conference on military technologies (ICMT) , organization IEEE . pp. pages 1--5

    2019

  27. [35]

    , author Wu, Q

    author Osco, L.P. , author Wu, Q. , author de Lemos, E.L. , author Gon c alves, W.N. , author Ramos, A.P.M. , author Li, J. , author Junior, J.M. , year 2023 . title The segment anything model (sam) for remote sensing applications: From zero to one shot . journal International...

    2023

  28. [36]

    , author Li, Z

    author Pei, D. , author Li, Z. , author Ji, R. , author Sun, F. , year 2014 . title Efficient semantic image segmentation with multi-class ranking prior . journal Computer Vision and Image Understanding volume 120 , pages 81--90

    2014

  29. [37]

    , author Zhaokui, W

    author Qingyun, F. , author Zhaokui, W. , year 2022 . title Cross-modality attentive feature fusion for object detection in multispectral remote sensing imagery . journal Pattern Recognition volume 130 , pages 108786

    2022

  30. [38]

    , author Liu, B

    author Ren, B. , author Liu, B. , author Hou, B. , author Wang, Z. , author Yang, C. , author Jiao, L. , year 2024 . title Swintfnet: Dual-stream transformer with cross attention fusion for land cover classification . journal IEEE Geoscience and Remote Sensing Letters

    2024

  31. [39]

    , author Ma, S

    author Ren, B. , author Ma, S. , author Hou, B. , author Hong, D. , author Chanussot, J. , author Wang, J. , author Jiao, L. , year 2022 . title A dual-stream high resolution network: Deep fusion of gf-2 and gf-3 data for land cover classification . journal International Journ...

    2022

  32. [40]

    , author Barbato, F

    author Rizzoli, G. , author Barbato, F. , author Zanuttigh, P. , year 2022 . title Multimodal semantic segmentation in autonomous driving: A review of current approaches and future perspectives . journal Technologies volume 10 , pages 90

    2022

  33. [41]

    , author Ghimire, B

    author Rodriguez-Galiano, V.F. , author Ghimire, B. , author Rogan, J. , author Chica-Olmo, M. , author Rigol-Sanchez, J.P. , year 2012 . title An assessment of the effectiveness of a random forest classifier for land-cover classification . journal ISPRS journal of photogramme...

    2012

  34. [42]

    , year 2013

    author Stone, J.V. , year 2013 . title Bayes' rule: a tutorial introduction to bayesian analysis

    2013

  35. [43]

    , author Chen, B

    author Tu, Y. , author Chen, B. , author Zhang, T. , author Xu, B. , year 2020 . title Regional mapping of essential urban land use categories in china: A segmentation-based approach . journal Remote Sensing volume 12 , pages 1058

    2020

  36. [44]

    , author Debnath, S

    author Woo, S. , author Debnath, S. , author Hu, R. , author Chen, X. , author Liu, Z. , author Kweon, I.S. , author Xie, S. , year 2023 . title Convnext v2: Co-designing and scaling convnets with masked autoencoders , in: booktitle Proceedings of the IEEE/CVF Conference on Co...

    2023

  37. [45]

    , author Park, J

    author Woo, S. , author Park, J. , author Lee, J.Y. , author Kweon, I.S. , year 2018 . title Cbam: Convolutional block attention module , in: booktitle Proceedings of the European conference on computer vision (ECCV) , pp. pages 3--19

    2018

  38. [46]

    , author Guo, S

    author Wu, W. , author Guo, S. , author Shao, Z. , author Li, D. , year 2023 . title Crofusenet: A semantic segmentation network for urban impervious surface extraction based on cross fusion of optical and sar images . journal IEEE Journal of Selected Topics in Applied Earth O...

    2023

  39. [47]

    , author Liu, Y

    author Xiao, T. , author Liu, Y. , author Zhou, B. , author Jiang, Y. , author Sun, J. , year 2018 . title Unified perceptual parsing for scene understanding , in: booktitle Proceedings of the European conference on computer vision (ECCV) , pp. pages 418--434

    2018

  40. [48]

    , author Shi, Y

    author Xu, F. , author Shi, Y. , author Ebel, P. , author Yang, W. , author Zhu, X.X. , year 2023 a. title High-resolution cloud removal with multi-modal and multi-resolution data fusion: A new baseline and benchmark . journal arXiv preprint arXiv:2301.03432

    2023 arXiv

  41. [49]

    , author Shi, Y

    author Xu, F. , author Shi, Y. , author Yang, W. , author Zhu, X. , year 2023 b. title Multi-modal multi-task learning for semantic segmentation of land cover under cloudy conditions , in: booktitle IGARSS 2023-2023 IEEE International Geoscience and Remote Sensing Symposium , ...

    2023

  42. [50]

    , author Lu, K

    author Xu, J. , author Lu, K. , author Wang, H. , year 2021 . title Attention fusion network for multi-spectral semantic segmentation . journal Pattern Recognition Letters volume 146 , pages 179--184

    2021

  43. [51]

    , author Rottensteiner, F

    author Yang, C. , author Rottensteiner, F. , author Heipke, C. , year 2018 . title Classification of land cover and land use based on convolutional neural networks . journal ISPRS annals of the photogrammetry, remote sensing and spatial information sciences volume 4 , pages 251--258

    2018

  44. [52]

    , author Ji, S

    author Yu, D. , author Ji, S. , year 2023 . title Long-range correlation supervision for land-cover classification from remote sensing images . journal IEEE Transactions on Geoscience and Remote Sensing

    2023

  45. [53]

    , author Liu, H

    author Zhang, J. , author Liu, H. , author Yang, K. , author Hu, X. , author Liu, R. , author Stiefelhagen, R. , year 2023 a. title Cmx: Cross-modal fusion for rgb-x semantic segmentation with transformers . journal IEEE Transactions on Intelligent Transportation Systems

    2023

  46. [54]

    , author Tang, X

    author Zhang, R. , author Tang, X. , author You, S. , author Duan, K. , author Xiang, H. , author Luo, H. , year 2020 . title A novel feature-level fusion framework using optical and sar remote sensing images for land use/land cover (lulc) classification in cloudy mountainous ...

    2020

  47. [55]

    , author Liu, B

    author Zhang, Z. , author Liu, B. , author Li, Y. , year 2023 b. title Fursformer: Semantic segmentation network for remote sensing images with fused heterogeneous features . journal Electronics volume 12 , pages 3113

    2023

  48. [56]

    , author Tu, K

    author Zhao, S. , author Tu, K. , author Ye, S. , author Tang, H. , author Hu, Y. , author Xie, C. , year 2023 . title Land use and land cover classification meets deep learning: A review . journal Sensors volume 23 , pages 8966

    2023

  49. [57]

    , author Persello, C

    author Zhou, W. , author Persello, C. , author Li, M. , author Stein, A. , year 2023 . title Building use and mixed-use classification with a transformer-based network fusing satellite images and geospatial textual information . journal Remote Sensing of Environment volume 297...

    2023

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.